Variability in particulate concentrations in a horse training barn over time



Reasons for performing study

Exposure of horses to airborne particulates during stable confinement has been linked with airway inflammation in these animals. Understanding that link requires accurate measures of exposures and greater understanding of the sources of variability in these exposures.


Area and breathing zone particulate concentrations were measured over time in order to determine the relative variability introduced by daily, monthly or between horse variations. Additionally, the relationship between area and breathing zone respirable particulate concentrations was examined.


The study was conducted in a Thoroughbred training stable. Breathing zone and area respirable particulate concentrations were measured over a 30-month period. Mixed-model analysis of variance was used to determine effect of month and year at the time of sampling and the daily variance upon area particulate concentrations. The effects of hay feeding method and horse variance on breathing zone measures were included in the model. Real-time concentrations of particulate matter with an aerodynamic diameter of 10 μm or smaller (PM10) were measured to determine the effect of barn door position. Significance was set at P<0.05.


Average area particulate concentration varied with month and year of sampling but daily variation was not significant. Maximum area respirable particulate concentrations were significantly affected by daily variation. Opening barn doors resulted in lower PM10 levels. Horses fed from hay nets were exposed to significantly higher concentrations of respirable particulates in their breathing zone than when fed hay on the ground. Horse-to-horse variability was significant. Breathing zone concentrations were significantly greater than area concentrations and the 2 measurements were not correlated.


While area respirable particulate concentrations reflected seasonal changes, these measures are poor predictors of individual horse exposure. Instead, methods of feeding and individual horse behaviour are important determinants of exposure.

Potential relevance

Studies investigating the effect of natural exposures on lung health in horses should consider the effects of individual behaviour and management practices on breathing zone exposure.


Inhalation of airborne particulates is an important cause of airway inflammation in the horse [1-4]. While recurrent airway obstruction represents the pulmonary response to particulates at its most severe, exposure to particulates is also implicated in the aetiology of inflammatory airway disease. Experimentally, inhalation challenge with organic particulates induces airway inflammation in otherwise healthy horses [3, 4]. Conventional stabling has been demonstrated to result in higher particulate exposures than pasture turnout [5] and stabling previously pastured healthy horses is sufficient to induce lower airway inflammation [6]. The prevalence of inflammatory airway disease across breeds and disciplines [7-9] indicates the universal importance of air quality in horse barns.

The impact of particulate levels upon respiratory health is partially determined by the size of the particulates. The American Conference of Governmental Industrial Hygienists describes 3 fractions for health-related aerosol particulate sampling: the inhalable, thoracic and respirable fractions. The respirable fraction, defined nominally as particulates with aerodynamic diameter of 4 μm or smaller [10], are particles small enough to penetrate the gas exchange regions of the human lung. This fraction has been considered an appropriate indicator of equine health risk when evaluating particulate levels in horse stables [11-14]. Closely matching the thoracic fraction, particulate matter with aerodynamic diameter of 10 μm or smaller (PM10) approximates those particles able to penetrate the thoracic conducting airways [10]. This measurement has also been used to assess air quality in horse stables [14, 15].

Particulate levels vary depending on numerous factors including ventilation, management and activity in the barn [5, 11, 12, 14]. Measures of airborne particulates also vary with sampling method. Breathing zone particulate concentrations provide the most accurate estimation of true exposure as the sampler is placed near the nostrils of the subject. In horses, this method of sampling is labour-intensive and can be limited by practical constraints such as subject compliance and safety. Most often, breathing zone measurements are made using integrated filter sampling methods and the use of sophisticated real-time particulate monitoring instruments is limited to research settings [12]. Conversely, area measurements obtained using stationary air sampling equipment allow continuous monitoring of particulate levels in the barn and permit the use of real-time monitoring equipment. However, stationary monitors underestimate true personal exposure, neglecting the ‘personal cloud’ of particulates generated by activity [12, 16-18].

Particulate levels can also be impacted by seasonal changes. Attempts to characterise seasonal changes in air quality of horse barns frequently rely upon measures of air quality over the course of 24–48 h. Such isolated measures may not represent particulate concentrations throughout the season due to high variability in these measures [19]. No published data are available to evaluate the day-to-day variability in particulate levels relative to seasonal changes in equine stables. Therefore, the purpose of this study was to measure respirable particulate concentrations over the course of 30 months at a single training barn, testing the following 4 hypotheses:

  1. Area respirable particulate levels will have significant day-to-day variability.
  2. Area respirable particulate levels will vary month to month and seasonal changes will emerge from this variability.
  3. Area PM10 concentrations will vary with external door position and month at the time of sampling.
  4. Respirable particulate concentrations at the breathing zone of stabled horses will be greater than area particulate levels and the 2 measurements will not be correlated.

Materials and methods

Description of facility (Fig 1)

Figure 1.

Schematic of barn layout and sampling locations.

A large, 36 stall dirt floor pole barn with 8 stalls along the east and west walls and a 12 stall addition at the northwest corner was used in this study. A further 8 stalls with overhead hay storage were located in the centre of the main barn along with a hot walker. The aisle surrounding the central stalls was used for jogging horses under saddle during inclement weather. In the main barn, the front wall of the stalls consisted of wooden boards extending roughly 1.2 m and the wall was completed with chain link fencing wire. A 1 m yoke style grill gate mounted roughly 0.5 m above the barn floor served as a door for each stall. Two sliding double doors, 3.6 m in height on the north and south walls provided access to the main barn. The 12 stall addition consisted of a single aisle, roughly 3.6 m in width and 20 m in length, with 6 stalls along each side of the aisle. The fronts of these 12 stalls were roughly 2.7 m in height, with solid boards reaching 1.4 m in height. The remainder of the stall front consisted of evenly spaced iron bars, with sliding stall doors constructed similarly. The area above the stall partitions was open. The barn ceiling of the addition, reaching a height of 4.2 m at its centre peak, was lined by foil-backed insulation material. The barn addition opened to the outdoors at the north end by way of a 3.6 m split sliding door.

All natural ventilation was provided by the 3 external sliding doors. When temperatures exceeded 27°C, an exhaust fan over the addition's external sliding door and box fans over each stall front provided supplemental ventilation. Each external sliding door remained open unless outdoor temperatures fell below -6°C or if inclement weather conditions occurred. All horses entered and exited the barn through the addition.

Throughout the barn, the floor consisted of packed dirt and the aisle floor was raked twice daily. Each stall was cleaned and bedded with sawdust once daily. Typically, all 36 stalls were occupied. Horses were fed alfalfa hay and oats rolled in molasses at approximately 07.00 h and 16.00 h daily. Each horse was exercised daily during this time window, 6 days per week. When not exercising, each horse remained in a stall. Horses were not exercised on Sunday.

Real-time area respirable particulate measurements

Real-time aerosol sampling was conducted with a Model 8520 DustTrak1 aerosol photometer with a 10 mm Dorr-Oliver nylon cyclone that samples particles with a 50% cut-off of 4 μm (respirable fraction). The flow rate was set at 1.7 l/min according to the manufacturer's recommendation and confirmed prior to each sampling period with a flow meter. Data were recorded at intervals of 20 s, each reading being an average of the values measured over each of the previous 20 s. Sampling was conducted for 24 h for 6–7 days during each sampling period. The DustTrak was positioned at the junction of the original pole barn and the barn addition during respirable particulate sampling (Fig 1).

Real-time area PM10 particulate measurements

Real-time aerosol sampling was conducted with the same Model 8520 DustTrak1aerosol photometer using the PM10 intake nozzle that samples particles with a 50% cut-off of 10 μm to examine the effect of external door position (open or closed). The flow rate was set at 1.7 l/min according to the manufacturer's recommendation and confirmed prior to each sampling period with a flow meter. Data were recorded at intervals of 5 s, each reading providing an average of values measured over each of the previous 5 s. Sampling was conducted for 5 min at each sampling location for each door position. Five sampling locations were used, with Tygon tubing connected to the inlet nozzle used to sample the air at the stall fronts indicated in Fig 1 (Nos. 1–5). Horses occupied the sampled stalls during each sampling period. Each day of PM10 sampling, the door position was randomly determined (all external sliding doors open or closed) and sampling conducted successively at each of the 5 stations. The door positions were then reversed and 10 min were allowed to elapse in order for particulate levels to reach steady state (data not shown). Sampling was then repeated at each of the 5 stations. This protocol was performed between 09.45 h and 12.15 h for 5 days during January and repeated for 5 days in March.

Breathing zone respirable particulate measurements

As part of a larger study, 12–36 month old Thoroughbred horses entering race training at the facility were enrolled. Air sampling was performed in the breathing zone of each horse over the course of 4–6 h between 10.00 h and 16.00 h once a week for 4 weeks. Gravimetric filter sampling was conducted with personal samplers (AirCheck 2000)2. The respirable fraction (50% cut-off of 4 μm) was collected using an aluminum cyclone2 with a flow rate of 2.5 l/min, per manufacturer's instructions. Pumps were calibrated before and after sampling (Defender Bios calibrator)2. Respirable samples were collected on 37 mm type AE glass fibre filters. The cyclone was secured to the noseband of the halter in order to sample dust at the breathing zone of the horse. The pump was secured to a surcingle placed around the girth of the horse and the pump connected to the cyclone with Tygon tubing secured to the mane and forelock of the horse. The horse was free to move around the stall as usual. Whether the horse was fed from a hay net or from the ground was recorded. Dust measurements were determined gravimetrically by subtracting the average of 3 tare weights taken before sampling from the average of 3 total weights obtained after sampling. Filters were placed in a desiccator for 24 h prior to weight measurements. Field blanks were prepared in an identical fashion and transported to and from the site of sampling. The Purdue Animal Care and Use Committee approved all procedures and owners signed informed consent forms prior to enrolling their horse in the study.

Statistical methods

Respirable particulate concentrations were summarised as mean (± s.d.), median, geometric mean and range. Wilcoxon rank-sum tests were used to compare breathing zone respirable particulate concentrations between horses fed hay in a hay net or from the ground, area PM10 concentrations between open or closed door positions and average respirable particulate concentrations between area and breathing zone measurement methods. A repeated analysis of variance (ANOVA) mixed model was performed with month and year as fixed categorical variables, day as a random variable and area (average and maximum) and breathing zone respirable particulate concentrations as outcomes. In the case of breathing zone measures, use of hay net as a fixed variable and horse as a random variable were included in the model. Significance levels for pairwise comparisons of factors means were controlled using Tukey's method for post hoc analysis. Simple linear regression was used to evaluate the relationship between respirable particulate concentrations and daily maximum temperature and relative humidity. Data were transformed in order to satisfy model assumptions: real-time area and breathing zone average particulate concentrations required log transformation while the inverse square root of real-time area maximum respirable concentrations was needed. Significance level was set at P<0.05.


Data were collected between May 2009 and December 2011. Particulate concentrations are summarised in Table 1 and Figure 2.

Figure 2.

Scatter plot of particulate concentrations vs. time.

Table 1. Respirable particulate concentrations (mg/m3) measured in the aisle of the barn (area measurements) or close to the horses' nostrils (breathing zone measurements)
 Area averageArea maximumBreathing zone average (no hay net)Breathing zone average (hay net)
  1. ND = not detectable; Real-time limit of detection = 0.001 mg/m3; Breathing zone limit of detection = 0.03 mg/m3.
Mean ± s.d.0.031 ± 0.0190.24 ± 0.360.097 ± 0.200.46 ± 0.65
Geometric mean0.0240.150.0580.21

Area particulate concentrations

Real-time area respirable particulate measures were obtained for a total of 257 days, with data available for February and May through November. Data were not available for January, March, April or December due to maintenance and unavailability of equipment.

Mixed-model analysis of log transformed area daily average respirable particulate concentrations revealed no significant day-to-day variability. Instead, the fixed effects of both month and year were significant, with area average respirable particulate concentrations in 2009 significantly less than those in 2010 (P = 0.0002) and 2011 (P<0.0001). There was no significant difference between 2010 and 2011 (P = 0.61). Respirable particulate concentrations were lowest in February and highest in August (Fig 3).

Figure 3.

Box-and-whisker plot of average area respirable particulate concentration by month. + = mean concentration; central horizontal line = median concentration; box = interquartile range; whiskers = maximum and minimum concentrations. Different letters denote statistically significant differences between groups at P<0.05.

Linear regression analysis of log transformed area average respirable particulate concentrations revealed a significant relationship with outdoor temperature (P<0.0001, r2 = 0.34, Fig 4) but not relative humidity (P = 0.87). The resulting fitted regression equation is given:

display math
Figure 4.

Scatter plot of log transformed area average respirable particulate concentrations vs. ambient temperature with regression line.

Mixed-model analysis of transformed maximum respirable particulate concentrations revealed significant day-to-day variation (P = 0.0083), but no effect of month at time of sampling (P = 0.18). Covariance estimates indicated that 26% of the random variation in maximum respirable particulate concentrations was due to daily variability. Maximum respirable particulate concentration measurements in 2009 were significantly less than 2010 (P = 0.0049) and 2011 (P = 0.0002).

Real-time area PM10 sampling was conducted between 09.45 h and 12.15 h between January and March 2010 on 10 separate days, for a total of 5 PM10 measurements per sampling station in late January and 5 PM10 measurements per sampling station in March for each door position. Barn temperature during the January sampling period (1.6 ± 4.1°C) was significantly lower than in March (10.2 ± 0.7°C). PM10 was significantly greater during the first sampling period when compared with the second period (P<0.0001). During the January sampling period, closing the external doors resulted in a 25% increase in PM10 regardless of sampling location (P = 0.012). Similarly, during the March sampling period, reducing ventilation by closing the doors resulted in a 30% increase in PM10 (P<0.001).

Breathing zone respirable particulate concentrations

Breathing zone particulate measures were obtained for 61 days, with data available for January and May through December (Fig 2). No breathing zone data were available for the months of February, March, or April due to lack of horses eligible for enrollment. Forty-one horses were sampled, for a total of 125 data points.

Breathing zone respirable particulate concentrations were significantly greater when horses were fed from hay nets than when hay was fed from the ground (P<0.0001, Table 1). Covariance parameter estimates indicated that between horse variability accounted for 24% of the random variation in the data (P = 0.037). Year and month at the time of sampling did not affect breathing zone measures, nor did day-to-day variability. Feeding hay from a hay net remained highly significant in the full mixed model (P<0.0001). Gravimetric analysis of field blanks confirmed that no detectable particle collection occurred when cyclones were merely transported to and from the farm without being connected to the air sampling pump.

Comparison of area and breathing zone average respirable particulate concentrations

When area average respirable particulate concentrations were compared with breathing zone measures obtained on the same day, concentrations measured at the breathing zone were significantly greater than area concentrations (P<0.0001) and the 2 measurements were not correlated regardless of the method by which hay was fed (Fig 5).

Figure 5.

Scatter plot of breathing zone respirable particulate concentration by feeding method vs. area average concentration.


Real-time area particulate concentrations

The mixed model was used to investigate the relative contribution of day and month at the time of sampling to particulate concentration variability. The use of season as a categorical variable was purposefully avoided in an effort to minimise misclassification, as weather patterns rarely follow the strict definitions of season by date. We observed that average area particulate concentrations did not change significantly day-to-day but rather were influenced by month and year of sampling, exhibiting a positive correlation with external temperature. Area respirable particulate concentrations were lowest in the winter and reached peak levels during the hottest summer months. This pattern is contrary to the conventional assumption that particulate concentrations in horse stables are greater during the colder months. However, the current study was limited by the scarcity of data points during the winter months, with area particulate measures only available for February 2010. Nevertheless, this February data set represented 7 consecutive days of sampling and thus provided more continuous data with which to evaluate the season than previously reported [14, 20, 21]. Furthermore, this relationship was confirmed by the positive correlation between area respirable particulate concentrations and temperature, with 34% of the variability in dust concentration explained by changes in ambient temperature. These findings are in agreement with Rosenthal and colleagues' observation that airborne particulates within the barn seemed to be weighted towards the respirable fraction in the summer months [20]. The hypothesis that rate of ventilation and temperature can exert differing effects upon particulate concentration depending on the particle size fractions in question is further supported by PM10 measurements made in this study. PM10 decreased with both increasing temperature (March vs. January) and increasing ventilation (doors open vs. doors closed). Other evidence for this phenomenon exists in the literature. Particle mapping in 3 racing stables at an American racetrack revealed that the number of airborne particles less than 5 μm was significantly lower in November than in July, while PM10 concentrations were significantly greater [14]. Inhalable dust (<100 μm) concentrations measured at 15 different equine properties between March and July were suggestive of a seasonal pattern, with lower concentrations detected during the summer months but no statistical analysis of this trend was reported [19]. Conversely, Riihimäki and colleagues reported that both total and respirable dust concentrations were greatest during the winter sampling periods but statistical analysis was not provided [21]. In addition, these measures cannot be directly compared with the current study, as concentrations were determined from a combination of stationary area sampling and barn worker breathing zone sampling. Unfortunately, no outdoor respirable particulate measurements were made in the current study. McGorum and colleagues have established that both respirable and total particulate levels measured in the breathing zone at pasture are significantly less than those measured in conventionally managed and low dust stables [5]. The effect of temperature or season on the magnitude of this difference has not been studied.

While the data presented here do not support the hypothesis that significant daily variations in the average area respirable particulate concentration exist, day-to-day variability in maximum area respirable particulate concentrations was significant. This daily variability is not surprising considering that daily maximums represent one measurement per sampling period that may be largely determined by a specific activity or event, for example raking of the aisle or horse cast in nearby stall. In contrast, daily averages are constructed from a large number of measurements logged every 20 s throughout the sampling period and are more likely to reflect the effects of temperature and ventilation in the barn.

In the current study, the mean area average respirable particulate concentration of 0.0306 mg/m3 was much lower than levels reported by Rosenthal and colleagues under natural conditions at 12 Indiana horse barns, which ranged from 0.10 mg/m3 in the winter to 0.20 mg/m3 in the summer [20]. This difference may be partially explained by differences in sampling equipment. In the current study, real-time measurement of area respirable particulate concentrations allows a lower limit of detection than the integrated filter sampling used in the previous study. Similarly, the geometric mean of average respirable particulate concentration in stalls under low dust conditions (0.20 mg/m3) reported by Woods and colleagues exceeds those reported here [11]. However, direct comparison is not appropriate as measurements were taken within the stall in the study by Woods and colleagues, whereas in the current study area measures were taken in the barn aisle. The median average respirable particulate concentration reported here, 0.027 mg/m3, falls between those reported for a low dust management system with open windows (0.018 mg/m3) and a conventionally managed stable with closed windows (0.096 mg/m3) [13].

Similarly, the median daily maximum respirable particulate concentration identified in the current study, 0.12 mg/m3, is greater than that measured under experimental conditions with biocompost bedding (0.08 mg/m3) but less than that reported when wood shavings were used (0.20 mg/m3) [22]. The mean maximum respirable particulate concentration reported for an experimental low dust management system using wood shavings bedding and haylage (0.22 mg/m3) is closely comparable with the levels reported here [12].

Breathing zone particulate concentrations

Breathing zone respirable particulate concentrations were highest when horses were fed from a hay net and there was significant variability among horses. The finding that feeding from a hay net results in greater than a 4-fold increase in exposure to respirable particulates compared with feeding from the ground is not surprising but illustrates that small differences in management can strongly influence exposure. Hay is a major source of particulates [12] and horses eating from hay nets tend to bury their muzzles into the hay, positioning their nostrils in close proximity to the particles liberated as the hay is agitated. Conversely, horses eating hay from the ground have their nares positioned above the source of particles, allowing some gravitational settling and diffusion of particles to occur, thereby reducing exposure.

As expected, breathing zone measures of respirable particulates were not affected by month at the time of sampling and did not display any evidence of seasonal variability. The significant variation explained by the random factor of horse has important implications. To the authors' knowledge, there is no other published evidence that individual horse behaviour can impact exposure. This evidence also supports the importance of an ‘equine personal cloud’ effect, a concept that is well established in human literature [16, 17]. The term ‘personal cloud’ is used to describe a microenvironment immediately surrounding the individual wherein particulate concentrations are determined by ambient particulate levels and particulates generated or resuspended by personal activity. In the case of the horse, these activities are certain to include eating hay, rolling and any other movement likely to generate airborne particles or cause settled particles to re-entrain. Therefore, personal sampling to measure particulate concentrations within the breathing zone provides the most accurate estimation of true exposure. The poor correlation between area and breathing zone respirable particulate concentration confirms that use of area measures will consistently result in underestimation of exposure, will be vulnerable to bias by ambient temperature and ultimately prevent the detection of individual differences in exposure.

When compared with the results of the current study, the mean breathing zone respirable particulate concentration of 0.22 mg/m3 reported when horses were experimentally managed under low dust conditions [5] is greater than that measured when hay was fed from the ground (0.097 mg/m3) but less than that when a hay net was used (0.46 mg/m3). The mean and median respirable particulate concentrations measured in the current study when hay was fed from the ground are comparable to those reported by others under conventional management systems (0.087 and 0.068 mg/m3, respectively) [13].

In the current study, all measures of respirable particulate concentration were made in a single equine facility. It will be important to repeat this analysis in other barns and in other climates to determine the generalisability of the results and to evaluate the impact of other factors such as barn construction, proximity to particulate matter sources and type of forage fed. In addition, other irritants such as airborne endotoxin and gaseous ammonia should be considered when evaluating the air quality in any horse barn.

Finally, the importance of any changes in air quality can only be determined if the relationship between exposure and respiratory health is better characterised. This will require not only accurate measures of exposure (i.e. breathing zone) but also simultaneous, reliable and sensitive indicators of lung function and airway inflammation. Additionally, it is not appropriate to assume that horses under similar management or even horses in the same barn, will experience similar exposures. Since method by which hay is fed and individual horse behaviour are major causes of variation in exposure, these factors need to be accounted for in future studies that explore the effect of exposure on respiratory health in horses.

Authors' declaration of interests

No competing interests have been declared.

Source of funding

Supported in part by the state of Indiana and the Purdue University School of Veterinary Medicine research account funded by the total wager tax.


Sincere thanks to Ms Donna Griffey for her technical assistance, Mr Xiaosu Tong and Dr Yu Michael Zhu for their statistical advice and Mr James Elliott and Mr Jeremy Stahley.


Drs. Ivester, Smith, Zimmerman, and Couetil contributed to study design. Drs. Ivester and Smith were responsible for data collection and study execution. Data analysis was performed by Drs. Ivester and Moore. Drs. Ivester, Zimmerman, Moore, and Couetil contributed to data interpretation. Dr. Ivester prepared the manuscript, which was edited by all authors.

Manufacturers' addresses

  1. 1

    TSI, Inc., Shoreview, Minnesota.

  2. 2

    SKC, Inc., Eighty Four, Pennsylvania.